Genetic variation in retinal vascular patterning predicts variation in pial collateral extent and stroke severity

The presence of a native collateral circulation in tissues lessens injury in occlusive vascular diseases. However, differences in genetic background cause wide variation in collateral number and diameter in mice, resulting in large variation in protection. Indirect estimates of collateral perfusion suggest that wide variation also exists in humans. Unfortunately, methods used to obtain these estimates are invasive and not widely available. We sought to determine whether differences in genetic background in mice result in variation in branch patterning of the retinal arterial circulation, and whether these differences predict strain-dependent differences in pial collateral extent and severity of ischemic stroke. Retinal patterning metrics, collateral extent, and infarct volume were obtained for 10 strains known to differ widely in collateral extent. Multivariate regression was conducted, and model performance was assessed using K-fold cross-validation. Twenty-one metrics varied with strain (p<0.01). Ten metrics (e.g., bifurcation angle, lacunarity, optimality) predicted collateral number and diameter across seven regression models, with the best model closely predicting (p<0.0001) number (±1.2-3.4 collaterals, K-fold R2=0.83-0.98), diameter (±1.2-1.9 microm, R2=0.73-0.88), and infarct volume (±5.1 mm3, R2=0.85-0.87). An analogous set of the most predictive metrics, obtained for the middle cerebral artery (MCA) tree in a subset of the above strains, also predicted (p<0.0001) collateral number (±3.3 collaterals, K-fold R2=0.78) and diameter (±1.6 microm, R2=0.86). Thus, differences in arterial branch patterning in the retina and the MCA trees are specified by genetic background and predict variation in collateral extent and stroke severity. If also true in human, and since genetic variation in cerebral collaterals extends to other tissues at least in mice, a similar "retinal predictor index" could serve as a non- or minimally invasive biomarker for collateral extent in brain and other tissues. This could aid prediction of severity of tissue injury in the event of an occlusive event or development of obstructive disease and in patient stratification for treatment options and clinical studies.